Recurrent conditional heteroskedasticity

We propose a new class of financial volatility models, which we call the REcurrent Conditional Heteroskedastic (RECH) to improve both in-sample analysis and out-of-sample forecast performance traditional conditional heteroskedastic models. In particular, incorporate auxiliary deterministic processes, governed by recurrent neural networks, into variance e.g. GARCH-type flexibly capture dynamics underlying volatility. The RECH models can detect interesting effects in overlooked existing such as GARCH (Bollerslev, 1986), GJR (Glosten et al., 1993) EGARCH (Nelson, 1991). often have good forecasts while still explain well stylized facts retaining well-established structures econometric These properties are illustrated through simulation studies applications four real stock index datasets. An user-friendly software package together with examples reported paper available at this https URL.